Spatial-aware detection of copy number alterations from spatial transcriptomics using SpaCNA

Spatial transcriptomics (ST) profiles genome-wide gene expression while preserving spatial context, yet accurate detection of copy number alterations (CNAs) in tumor ST data remains challenging. Here, we present SpaCNA, a computational framework that integrates multi-modal information of ST for robust CNA detection. SpaCNA aggregates expression from neighboring spots with similar morphological features and leverages a hidden Markov random field model incorporating spatial continuity for reliable CNA detection in ST datasets. Further, SpaCNA can reconstruct 3D CNA profiles with spatial continuity across consecutive slices when applied to 3D ST datasets. Extensive benchmarking on simulated data and real cancer datasets demonstrates SpaCNA's superior accuracy, achieving up to 0.95 F1-score in CNA detection and tumor region identification. In applications to breast cancer and colorectal cancer, SpaCNA reveals tumor boundaries and spatially distinct subclones with context-dependent interactions within the microenvironment. Notably, SpaCNA performs CNA detection in a 3D ST dataset of head and neck squamous cell carcinoma, revealing the tumor evolution trajectory of three subclones in 3D space. By providing accurate CNA inference, SpaCNA facilitates the analysis of intratumoral heterogeneity and spatial cancer biology.